Disillusionment with centralized planning has resulted in a global shift towards “participatory” models of development. Anchored by marquee approaches such as Participatory Budgeting (hereafter PB) and Community-Driven Development (hereafter CDD), participation has become a ubiquitous tool for devolving decision-making to the community level (Baiocchi and Ganuza 2017; Casey 2018). The World Bank, in particular, has invested more than $85 billion USD in participatory programs, with participation widely integrated into policy and intervention design across both the Global North and South (Baiocchi and Ganuza 2017; Mansuri and Rao 2013).

Participatory planning is loosely defined as the direct and broad-based participation of communities in a public process for allocating public goods. Within this category of interventions, these programs have pledged a wide range of potential benefits, especially in “difficult” or “fragile” contexts (King et al. 2010). Participatory planning processes are believed to empower marginalized groups, foster democratic governance, and enhance overall welfare and public resource efficacy, all while nurturing social cohesion and social capital to promote future stability and reduce conflict (Mansuri and Rao 2013; King and Samii 2014).

Despite its widespread adoption, initial enthusiasm for participatory planning has been tempered by a number of ambiguous or null empirical findings regarding its social impact. Systematic reviews of participatory planning have consistently found modest positive economic welfare effects and increased access to public goods, but evidence on social impact—democratic governance, increased social capital, and equitable social norms—is markedly weaker and inconsistent (Casey 2018; Mansuri and Rao 2013; Touchton and Wampler 2023; White et al. 2018; Wong 2012).Footnote 1 These reviews have largely advanced the pessimistic conclusion that participatory planning does not “work” as a class of interventions with regard to social change or, at minimum, its operation may require supportive factors often absent in implementation contexts (Touchton and Wampler 2023).

Our study does not question the value of the many rigorous evaluations of CDD for determining the retrospective value of money spent on participatory programs. As systematic evaluations have pointed out, a diversity of sophisticated experimental and quasi-experimental studies have strong internal validity (Casey 2018). Nonetheless, we argue that prospective conclusions about whether participatory planning works as a category of intervention are premature in two major respects.

First, evaluating participatory programs using experimental and quasi-experimental designs may be valuable for retrospectively determining the value of money spent, but—unless accompanied by a mechanistic account explaining what occurs in the “black box”—it is of limited prospective value for predicting when and whether a program will produce similar outcomes in another context (Andrews et al. 2017; Cartwright et al. 2020; Cartwright and Hardie 2012; Woolcock 2013). While some systematic reviews recognize this limitation and have called for more “causal chain analysis” (e.g., White et al. 2018), the research in response has been limited, and there is no general agreement on where and why participatory planning interventions work or fail as a category.

Second, systematic reviews aggregate evidence on the assumption of input homogeneity within and between program evaluations. This assumption is typically violated, because participatory planning is a complex program, requiring significant adaptation between contexts. Under these conditions, aggregating existing evidence on participatory planning in systematic reviews will inevitably lead to a conclusion that the evidence is mixed or weak, because inputs are inconsistent. Such studies cannot explain variation retrospectively within or between programs or prospectively suggest new strategies for practitioners to improve. While existing systematic reviews have stressed the need for more causal chain analysis, they do not suggest nor confront the desirability of input heterogeneity within program implementation and rely on the aggregation of experimental evidence based on assumptions of homogeneity (King et al. 2010; White et al. 2018).

Thus far, the evaluation literature on participatory planning has taken what we describe as a “what works?” approach. Most of the existing literature has been framed as asking “Does participatory planning work?”—which implies a binary and universal answer. We seek to address the ambiguity and limitations of the existing literature on participatory planning by moving away from this animating question. Instead, we ask a slightly different question aimed to break the current deadlock in the literature and policy discussions on participatory planning: What aspects of intervention design and implementation shape how a given participatory planning intervention functions in a particular context?

Our approach steers practitioners and scholars towards a conditional understanding of when rather than whether participatory planning interventions “work” (Woolcock 2019). Such an understanding emphasizes that program designers and implementers should have significant agency to make adaptations and course corrections throughout. We acknowledge that customizing program design and encouraging space for adaptation and discretion at the ground level break foundational assumptions of input and dosage homogeneity required for randomized controlled trial evaluations. However, ignoring the existing reality of desirable input variation within and between programs in systematic reviews of participatory planning will inevitably lead to the conclusion that the evidence is mixed or weak, which cannot convincingly explain variation retrospectively within or between programs nor prospectively suggest new strategies for practitioners.Footnote 2

To pursue this question, we use process tracing case studies of one specific participatory planning intervention, Ward Development Planning (WDP), which was implemented in five counties in Northern Kenya from 2017 to 2022, as part of a wider 5-year USAID program. Our research was carried out in the final year of implementation, allowing us to understand the operation of the institutional model during implementation.Footnote 3 We use within-case and comparative analyses of 12 out of the 32 wards where WDP was implemented to collate the factors that allow or block causal processes (Beach and Pedersen 2019; Cartwright et al. 2020; Schmitt and Beach 2015). We use elements of theory-building process tracing (Kaidesoja 2019; Beach and Pedersen 2019) to start constructing an explanation of how and why participatory planning interventions do or do not work in practice, focusing on mechanisms which “shed light on how a cause (or set of causes) contributes to produce an outcome,” relying on “empirical material” to construct these plausible mechanisms (Beach and Pedersen 2019, 270).

This study is not evaluative, and we refrain from making claims regarding the realization of long-term outcomes, whether concerning social cohesion, accountability, or the transformation of social norms, as our fieldwork took place during the implementation phase. Instead, our focus is on understanding the circumstances under which a participatory development program functions in a manner consistent with theories of downstream outcomes. This emphasis on function mirrors a distinction drawn by Gibson and Woolcock (2008) between explaining the causes of empowerment and the effects of empowerment on subsequent development outcomes. Our paper primarily concerns the former—identifying the decisions and factors that can facilitate or impede the mechanisms through which inclusive participation leads to empowerment.

In other words, we aim to articulate the factors contributing to a successful “treatment” during program implementation rather than solely relying on the intent to treat. For instance, we argue that at the micro-level, discretionary adaptation by ground-level implementers is crucial for the program to foster empowerment, ensuring both broad participation and preventing elite capture. While many theories of participatory planning assume these factors, we contend that their successful operation depends on various contingencies across various levels.

Based on our analysis, the WDP program was contingently successful in shifting governance and social patterns across three levels: macro, meso, and micro (Fig. 1).Footnote 4

Fig. 1
figure 1

Three points of contingency for participatory planning functioning

At the macro-level, the WDP’s level of implementation encountered a genuine governance gap. As a ward-level institutional intervention, the WDP provided a needed ward-level political structure for collectively advocating to both the county and international donors. At the meso-level, the rule-based design of the intervention was matched with contextual factors, allowing them to operate. For example, the elected representatives of the Ward Planning Committee (WPC) were accountable to their local communities due to pre-existing social-embeddedness and village-level social cohesion. This aligned with the institutional design of WPC using villages or village clusters as constituency units. At the micro-level, ground-level implementers use local discretion to translate abstract intervention design into real-world action, a process which is unobservable to centralized management but repeatedly a key factor for the intervention to function as intended. Significant and discretionary adaptation during implementation required ground-level implementers to possess a combination of local knowledge, motivation, and an understanding (at least on an intuitive level) of the program’s theory of empowerment as requiring both broad participation and deliberation. In this, ground-level implementers functioned as “Habermasian bureaucrats” (Gibson and Woolcock 2008), presiding as custodians over the process of implementation to both generate and protect empowerment, by deciding when to adapt to or confront local norms or institutions.

This paper is in five sections. The “Literature Review” section situates our study within the literature. The “Methodology” section describes the research methodology. The “Context and Description of Intervention” section summarizes the Kenyan context and the WDP program. The “How and Why Participatory Planning Interventions Work in Practice” section presents analyses of our sub-national case studies. The “Conclusion and Implications” section concludes with suggestions for future research.

Literature Review

Existing Literature

The scale of participatory planning within the development sector has led to a significant body of evaluation and research. Systematic reviews of this evidence have predominantly been evaluative, seeking to gauge whether participatory programs “work” to achieve anticipated economic, governance, and social benefits outlined by funders and participants. The results of these syntheses have been pessimistic conclusions that the interventions produce modest welfare benefits while producing only mixed or null effects on social outcomes (Casey 2018; King and Samii 2014; Mansuri and Rao 2013; Touchton et al. 2022; White et al. 2018).

For these systematic reviews to be useful in providing meaningful prospective guidance to future interventions, we argue that they must first explain why programs fail. Critically, systematic reviews must be able to discern between a failure of implementation or process (broadly failure of treatment) and a failure of theory. Only the latter answers the question of whether participatory program “works.” Additionally, existing studies do not adequately address the significant input homogeneity of programs, threatening the generalizability of their claims.

While existing systematic reviews acknowledge the need for causal chain analysis to understand why evaluations point to mixed or null results, only King et al. (2010) and to a lesser extent White et al. (2018) provide substantial analysis across studies. These studies point to a number of possible reasons why programs fail, including insufficient participation whether due to a “funnel of attrition” or a failure of implementation. However, the authors of these studies acknowledge that their analysis is unable to disambiguate between a failure of implementation (inadequate treatment to induce substantive and broad-based participation) and a failure of theory (participation in local decision-making does not lead to empowerment) (King et al. 2010). The inability to distinguish between these two types of failure is a key weakness of a “what works” approach to studying complex programming, which our study attempts to address.

Other studies present conditional accounts for why participatory planning may fail in certain contexts. These studies include the possibility that communities have inadequate social cohesion to initiate collective action, suggesting program may use rather than create social cohesion or at least a threshold effect.Footnote 5 Another concern is diagnosis failure, where deployment of participatory planning is misaligned with the problems within communities. For instance, in Afghanistan, participatory planning institutions may have crowded out strong and operational informal institutions, ultimately reducing accountability (Beath et al. 2015; Murtazashvili 2016). A related diagnosis failure arising in the literature is that intra-community (bonding) social capital may not be the binding constraint for collective action in certain contexts, explaining why encouraging social interactions via participation has null effects on contributions to public goods and other downstream outcomes (Casey 2018).

An additional issue undercutting the attempts by systematic reviews and evidence synthesis to determine whether participatory planning “works” is a lack of attention to input homogeneity. Systematic reviews aim to aggregate evidence on a single intervention implemented across contexts to achieve externally valid and prospective claims. This aggregation requires the assumption of input homogeneity for all interventions included in such analysis, meaning they all represent common inputs and causal pathways. For participatory planning, this aggregation has occurred under marquee program categories, such as “Community-Driven Development” or “Participatory Budgeting.” Beyond the threat to input homogeneity, no reviews confront the tension between this assumption and the possibility of desirable input heterogeneity, where adaptations are needed for programs to operate (King et al. 2010; White et al. 2018).

However, participatory planning interventions are complex and composed of many interlinked activities and decisions that are both transaction-intensive and discretionary (Pritchett and Woolcock 2004).Footnote 6 Design elements include community mobilization, facilitation, trainings, engagement with government actors, generation and approval of community plans, size of block grant, and often, selection of delegated community representatives (Touchton et al. 2022; Mansuri and Rao 2013). None of these elements is standardized between instances of participatory planning, and both intensity of implementation and capacity of implementers represent additional sources of variation (Casey 2018). Based on evidence from the education literature, intervention outcomes may be highly sensitive to choices made during design and implementation (Angrist and Meager 2023; Kerwin and Thornton 2020; Pritchett 2017). Given that participatory planning is at least as complex as education, if not more so, this represents a serious barrier to external validity.

Looking beyond formal program design, a recurring finding in empirical studies of participatory planning programs (CDD and PB) is that local implementing actors have a wide de facto degree of discretion regarding the implementation of participatory planning programs. While program theories of change offer an abstract rationale concerning causal processes, this rationale functions merely as a basic framework (Baiocchi et al. 2011; White et al. 2018) or to a more abstract extent, a set of “values” (Touchton et al. 2022). Within this framework, the implementing agent makes numerous tactical decisions on how best to activate the causal logic envisioned in the program’s theory of change. This concept is closely related to the idea of “institutional bootstrapping,” as described by Baiocchi et al. (2011), but it has gained little traction in policy circles or systematic reviews, apart from non-theoretical appeals for adaptive approaches to implementation.

Our study also benefits substantially from excellent qualitative scholarship on participatory planning projects. We build directly on a number of insights from this literature, such as the importance of variation at the “community” level (Bebbington et al. 2004) especially during the contested “bricolage” processes of participatory planning implementation (Pain 2018); the uneven operation of a program’s implementation and theory of change due to cultural, institutional, political, and capacity constraints (Hanif et al. 2022); and the meaningful operation of informal institutions in the absence of a formal state (Murtazashvili 2016). Baiocchi et al. (2011) likewise present excellent case studies exploring for why institutional designs of a single program (PB) such as delegated versus direct participation, can and should be adapted between Brazilian municipalities, depending on contextual factors, such as the presences of strong and independent civil society. Perhaps most closely, our study builds on Barron et al. (2011) and Gibson and Woolcock’s (2008) explorations of CDD in Indonesia, which address two of the three levels of adaptation we identify: first, the importance of pre-existing contextual factors (e.g., local capacity for conflict resolution in their study) and second, the pivotal importance of allowing ground-level implementors to have reasonably wide discretion in participatory programs, operating as “Habermasian”—rather than Weberian—bureaucrats, who use local knowledge to “adjust to some local norms and resist others” (2011, 267).

Our study also speaks to critiques of randomized controlled trials (Bédécarrats et al. 2020). Within international development and public administration, scholars critical of randomized controlled trials have argued that organizational capacity and implementation are under-appreciated determinants of program success. Broadly described under the banner “doing development differently,” these scholars also emphasize that locally adaptive and iterative approaches are necessary when programs are transaction-intensive and require local discretion (Andrews et al. 2017; Levy 2014; Pritchett and Woolcock 2004). Others argue that allowing implementers to navigate by judgment often yields better results than managing for compliance or implementing more intrusive top-down accountability structures to ensure standardized implementation (Honig 2018; 2022).

Finally, while our re-framing of scholarly inquiry on participatory planning is not a defense of participatory planning per se, it offers a useful complement or counterpoint to existing defenses. Some scholars have argued that Community-Driven Development should be more tightly defined in terms of macro-elements—such as integration with other governance reforms and quality of administrative implementation, e.g., on-budget versus off-budget programs (Wong and Guggenheim 2018). In contrast, our reframing abandons the quest for greater input homogeneity for desirable input heterogeneity, understood as the contingent operation of theorized causal mechanisms, dependent on the interplay of inputs and context. This allows us to shift focus away from efforts to circumscribe what the “true” model of participatory development (or its sub-categories) is and instead focus on how programs can and should be best customized to unique national and sub-national contexts.

Another defense is that inflated, “panacea-like” expectations for participatory planning may lead to overly negative assessment of their impact, especially with regard to social outcomes. More realistic a priori expectations may allow us for a reassessment of interventions as successes—albeit, perhaps not in all respects (Samii 2023). Our work is complementary to this and other calls for reassessment of what CDD is and how it ought to be understood (Bennett and D’Onofrio 2015). However, our approach is distinct in rejecting the question of whether participation “works” as unduly constraining, and we emphasize that participatory planning is a flexible intervention which, in theory, may achieve a wide range of outcomes depending on the success or failure of adaptation across the three levels we identify.


Case Selection

The decision to study the WDP program was prompted by two indicators that suggested it was a case of a functional participatory planning intervention. Firstly, the implementing agencies, Mercy Corps and ACDI-Voca, alongside the funder, USAID, emphasized that the program had led to substantial new participation and community mobilization during its implementation. They highlighted initial indications of power dynamics shifting between wards and the county government (Mercy Corps and ACDI-Voca 2022). Secondly, the Frontier Counties Development Council (FCDC), a regional block comprising ten northern counties in Kenya, expressed interest in exploring the possibility of expanding the WDP model based on the experiences gained from piloting it in five of its ten member counties. Recognizing it as a vital mechanism for community empowerment, the FCDC committed to providing technical support to develop and pass bills in each respective county, aimed at institutionalizing the WDP through a Ward Development Fund.

During the reading of a bill in 2022 in the Turkana County Assembly, a county legislator favorably contrasted WDP’s potential for empowering communities in public planning and budgeting, vis-à-vis the status quo approaches to public participation:

“I know that, in many cases we have carried out participatory exercises… [however] most of the ideas and proposals of the communities have not been captured in the budget. But this bill will empower communities to identify their key priorities to be implemented [in the budget]. This bill will make us now enjoy the main reason of devolution. We will have, for the first time, our members of the public making key decisions for how they want their own finances to be utilized.”

(County Assembly of Turkana 2022)

Our three county-cases are drawn from the five Kenyan counties in which the WDP intervention was implemented. From these five, we conducted a case study of the WDP intervention in three counties, covering twelve wards: Garissa County (five wards, out of five total), Turkana County (three wards, out of nine total), and Isiolo County (four wards, out of six total).

Garissa, Turkana, and Isiolo were selected from among the five potential counties implementing the WDP program to maximize variation of ethnicity, religion, and geography. Isiolo is the most ethnically diverse, while Turkana and Garissa are largely homogenous. Turkana is predominantly Christian and Garissa Muslim, while Isiolo is religiously diverse. Geographically, Garissa sits on Kenya’s Eastern border, Isiolo is centrally located, and Turkana is located on the Northwest border.

Data Collection

Our case study is based on in-depth qualitative field research carried out in October and November 2021. Data collected includes 58 interviews and 7 focus group discussions, carried out with program participants (n = 54), county and national government stakeholders (n = 12), and implementing staff (n = 11) with breakdown in Table 1.

Table 1 Respondent breakdown by gender, age, affiliation, and county

Selection occurred through consultation with the program team in each county, both ground-level implementers and their county-level managers who held relationships with the WPC members and county government. This selection was deliberate, meant to maximize variation in the identities and roles of WPC members. This raises the risk of motivated selection by the NGO and desirability bias in responses. Additionally, when possible, we aimed to address motivated selection by requesting large percentages of WPC members be interviewed, limiting the ability to sort individuals. To limit desirability bias within interviews, it was made clear at the start of each interview that this research was fully separate from the implementation, non-evaluative, confidential, and unrelated to direct future benefit or punishment. Interviews were conducted in private spaces, outside the earshot of staff or other participants, and besides co-respondents in cases of FGDs.

The selection process for respondents evolved between counties in response to the team’s learning as the fieldwork progressed. In Garissa County, respondents were mostly the three WPC officers (treasurer, secretary, and chairperson), and they traveled to the central Garissa Town (the county capital) due to our limited time for travel. For Turkana County, we reversed the process to allow a wider range of respondents, traveling to outlying wards. We selected fewer wards to engage with (3) but expanded the number of respondents per ward to around 10. This was a majority of all WPC members, including officers and non-officers, and we prioritized one-on-one conversations with female members, an under-represented group in the Garissa interviews. In Isiolo County, we again traveled to outlying wards and interviewed roughly seven members per ward, again selecting across covariates of age and gender.

Process and Consent

Interviews with participants and government staff were semi-structured and lasted roughly 30–40 min. FGDs lasted 60–75 min and were used when there were multiple of the same category of respondents (e.g., elder male) or in two ad hoc cases, when a set of people from the same village needed to leave quickly.

Given many respondents spoke only local languages, a local translator, independent of the program implementation, was present for all interviews and FGDs. Participants who spoke English could select between English or the local language, according to their preference, with the possibility of switching within the interview for more complex discussions.

To promote honest responses and limit potential harm, participants and government respondents were informed about the purpose of the research and informed that their identities would be kept confidential in all external reports. All gave verbal consent to be interviewed and, separately, to be recorded. Respondents were given the opportunity to turn off recording to discuss sensitive topics. All were told that no benefits or punishment would relate to their participation, and no identifiable information would be communicated, including to NGO program staff.


All interviews were reconstructed using either voice recording (when recording was affirmatively accepted by the respondent) or detailed field-notes. These were then systematically coded and analyzed using NVivo. An initial set of eight interviews were used to create a base codebook, against which the remaining interviews and FGDs were then coded, though the codebook continued to evolve at a slower pace if new mechanisms were encountered. Analysis consisted of looking for the causal “fingerprints” of mechanisms operating within-case (Beach and Pedersen 2019) in addition to between-case patterns across locations, including both most common and deviant cases. Attention was also paid to corroboration of empirical accounts between different respondents and categories of respondents. Though not exhaustive due to space limitations, in the account below, we connect claims to an illustrative sub-set of the interviews and focus groups which they were drawn using footnotes to document internal citations with key identity characteristics (e.g., #—Turkana, female, youth).

Context and Description of Intervention

Context: Kenya’s ASAL Regions

Studying the application of a participatory planning approach across Kenya’s arid and semi-arid land (ASAL) region is relevant for both policy and theory. Since 2013, Kenya has decentralized its political landscape. This has provided a degree of self-governance to the northern ASAL regions which comprise our case studies. Despite accounting for over 90% of its landmass and 30% of its population, the ASAL region of Kenya has long been politically and economically marginalized by the central Kenyan government (Odhiambo 2013). It was only in the late 1990s and 2000s that the government explicitly noted the need for greater “equity” in socio-economic development, implicitly admitting historical neglect of investments in services, infrastructure, and economic growth policy.

Besides a shift in national policy, the 2010 constitution created a decentralized governance system, where key legislative and executive functions were devolved from the central government to newly created county-level governments across the country. Along with fiscal devolvement of 15% of national revenues, county-level governments were tasked with providing many public goods previously the responsibility of the national government, such as health care, roads, and early childhood education (Kimenyi 2013). This funding is allocated using 5-year County Integrated Development Plans (CIDP) and Annual Development Plans (ADP). Though the constitution mandates that CIDPs be informed by public participation, there are few details about what constitutes sufficient participation or how demands should be integrated into the CIDP.

Despite these imperfections, devolution represents a genuine sea-change in Kenya’s political structure. The creation of a legislative County Assembly and county governor, alongside an array of county-level departments and committees (e.g. county steering committee, county health department), created the first elected, sub-national institutional interface between citizens and their government. Previously, the primary government interface was an appointed village chief representing the national government. Within the county government, the constitution also generated a new administrative level: wards, which are the primary administrative districts under the county and the electoral districts for the Members of the County Assembly (MCAs). The county governor also appoints Ward Administrator (WA) for each ward, a position tasked with implementing county policy and facilitating county-services. There are no ward-level legislative or executive positions, and there has been little domestic or international attention to ward-level civic organizing or institutional capacity.

Strengthening Devolution and the Ward Development Plan Model

Devolution alone cannot ensure governments are more accountable. Indeed, sub-national governance may reproduce destructive behaviors at a smaller scale (Mueller 2019). In Kenya, devolution has dramatically improved the dispersion of development funding from a national perspective by directly allocating significant fiscal resources to the county-level. However, county-level politics still suffers from clientelism, corruption, disorganized civil society, and weak political parties (Cheeseman et al. 2020; Hassan 2020; Oloo 2020).

If increasing the proximity of government structures to the populace is insufficient to avoid reproducing negative governance patterns, how then should the Kenyan government and international community promote democracy and “good” governance in practice, especially at the local level?Footnote 7

A common response is to utilize participatory planning institutions to facilitate democratic or “good governance” practices at the subnational level. These institutions allow for the direct and official engagement of citizens in a public process, involving the identification of public policy challenges and the proposal of public projects to tackle these issues (Smoke 2008, as cited in Sheely 2015). In these community meetings, ordinary citizens engage in a deliberative process to collectively rank public goods and communicate those priorities to governments and development funders. These participatory planning interventions are meant to empower ordinary citizens and marginalized groups in the face of existing elite or parochial interests (Gibson and Woolcock 2008; Sheely 2015), while also improving the efficient targeting of public spending by drawing on local knowledge often lacking in higher-level government structures (Oates 1999, as cited in Casey 2018).

This report focuses on a particular participatory planning process known as the Ward Development Plan (WDP) model. This ongoing governance intervention is part of the Livestock Market Systems program funded by USAID from 2017 to 2022. The development and execution of the WDP process were spearheaded by Mercy Corps within the implementing consortium. The WDP model itself is a participatory planning process that aims to equip community members with the skills to assess their own requirements, strategize projects, and hold government and development actors accountable for development priorities and service delivery results.

The WDP planning process primarily revolves around mobilizing communities to identify development needs and priorities, formalizing these priorities in a Ward Development Plan, and electing a diverse Ward Planning Committee (WPC) spanning various villages, gender groups, and age brackets. The WPC is tasked with overseeing WDP implementation and engaging with county government, along with other local development actors at the ward and county levels.

The WDP model serves as an amalgamation of the two primary models of participatory planning: participatory budgeting, which originated in Brazilian municipalities and has since proliferated globally (Baiocchi and Ganuza 2017; Touchton et al. 2022), and Community-Driven Development, which gained prominence worldwide, particularly through the initiatives of the World Bank. Therefore, it serves as a valuable case study for drawing theoretical insights into how specific design elements of participatory planning processes function at a more detailed level, aligning with the “meso” level referenced throughout the rest of the paper (see Table 2).

Table 2 Comparing participatory planning models by design features

These design featuresFootnote 8 include the level of decision-making power, which can be binding (directly controls funding) or consultative (only advises the control of funding); the type of participation, which can be either direct (public fora for deliberative decision-making) or delegated (public selection of representative delegates who engage in deliberative decision-making); the funding source and process facilitator (government or NGO); and whether the funding source is specified (predetermined) or non-specified (undetermined).

Table 2 represents the canonical or dominant applications of each model, based on a review of the CDD (Casey 2018; King and Samii 2014; Mansuri and Rao 2013; White et al. 2018) and PB literatures (Touchton et al. 2022; Baiocchi et al. 2011; Baiocchi and Ganuza 2017; Wampler 2012). Table 2 is not meant to be comprehensive—there are infinite potential design variations in participatory planning, such as quotas based on identity groups, intensity or number of meetings, and community contribution requirements—and there will certainly be exceptions to even these minimal characterizations within each class of programs. However, this serves to reinforce our broader thesis, which is that none of these intervention classes can or should be treated as homogenous.

How and Why Participatory Planning Interventions Work in Practice

Macro: Filling a Genuine Governance Gap

One key factor influencing the functioning of participatory planning interventions is whether there exists a genuine gap in governance and collective action concerning the allocation of public resources (Fig. 2). We find that government and civic institutions at the ward level had weak infrastructure for collective action or advocacy, and therefore, the WDP process, and especially the Ward Planning Committee (WPC), was able to fill a sorely needed gap for community participation and input. The members of the WPC were able to fill this gap by acting as legitimate, ward-level representatives, and were able to foster wider communication between the county and the village-level communities throughout the ward.

Fig. 2
figure 2

Macro level: filling a genuine government gap

This weak capacity for collective action at the ward-level is most attributable to Kenya’s still early process of devolution, which only began implementation in 2013, following many decades of highly centralized governances. As these newly formed county institutions are still in the early stages of gaining experience and capacity, often with the help of technical advisors (Neely et al. 2021), far less attention has been paid to the potential for collective decision-making in the lower administrative level newly formed in devolution: the ward. The handful of early experiments with ward-level collective action have been limited to climate change mitigation alone rather than decision-making about public goods holistically, and implementation is uneven, not covering most wards in Kenya, such as those in our sample.Footnote 9

In the face of this gap, the WDP process produces a public good in the form of a holistic, 5-year ward-level development coordination document, meant to complement existing county-level planning documents over the same 5-year period. Both government actors and participants interviewed hoped that this document would ensure that county and NGO projects were better aligned with local priorities, as identified by the community. This is a finding held across all counties and was corroborated by diverse stakeholders.

Most Ward Administrators (WAs) felt that the WPC provided a coherent and legitimate intermediary for representing the ward population, gathering information about local needs, and, in some cases, directly allocating public good due to a better understanding of the “grassroots.” For instance, in Isiolo, some Ward Administrators devolved disbursement of tuition assistance to WPC members, who were better placed to direct it to the “most vulnerable” in their constituent village(s).Footnote 10 County-level officials similarly felt that, prior to the WDP process, they had to rely on MCAs and other individual personal relationships to gauge community needs at either the ward-level and below.Footnote 11 Corroborating this, WPC members noted a lack of political organization at the ward level, while simultaneously feeling that the county-level government—both the administration and their assembly representatives —were too often inaccessible or non-responsive.

The WPC also created a more institutionalized information flow, less reliant on personal connections to “get things done.” This helped dilute dependency on local-level actors with links with the county government (e.g., popular religious leader or well-connected elders).Footnote 12 This was particularly useful as distribution of connections was perceived as uneven, leaving some villages with multiple well-connected representatives and others with none or negative relationships.Footnote 13

Though MCAs were elected as county-level legislative representatives, they were perceived as unaccountable within their wards, in part due to being more concerned with political machinations in the county capital than representing their ward. In addition, MCAs were perceived as having stronger connections with some villages or social groups within their ward, based on pre-existing ethnic ties usually through older males.Footnote 14MCAs were also commonly perceived as only available during election season every 5 years, and MCA’s “home” village was often expected to be better represented within the ward.Footnote 15 In comparison, the WPC represented a more fine-grained layer of political representation, including providing an additional interface for their constituent communities to engage the MCA, who were ex officio members by default, alongside other county-level officials. WPCs also reported interfacing with national-level development funding via the National Government Constituencies Development Fund.Footnote 16

The WPC acted as an intermediary between the grass-roots and government, communicating the community’s “real” needs to both the county and NGOs. This reveals a previous information gap, in which the county was insufficiently aware of community-level priorities. Existing intermediaries, WAs and MCAs, acknowledged their inability to gather or transmit community-level details due to the size of wards. This lack of local information had material effects. Both NGO and government funding was reportedly often misprioritized, funding either lower priority projects or “white elephants.” White elephant projects were described as unusable (e.g., a borehole near salty water) or redundant (e.g., building a market too close to an existing one). In one shocking instance, a county government project was actively harmful to climate resilience, when the county placed a well in the middle of a strategic grazing reserve without consulting the community.Footnote 17 This drove foot traffic (and grazing) into an area the local pastoralist community had designated for emergencies only.

“The county built a second market next to a previously built one. Now it is used for shooting practice . No one uses the new market. It was a waste of money.”

16 - Turkana, Chief, Male, Adult

The WDP process is not the first participatory planning intervention instituted across Kenya. Indeed, Kenya’s 2010 constitution mandates some degree of public participation in county-level planning. While participation in county planning and budgets is mandated nationally, the quality of implementation is uneven, and the resources allocated vary widely. While all counties reported some degree of public consultation about upcoming county budgets in previous years, these public consultations were under-resourced and fleeting. In no cases did previous public consultations encountered in the counties where WDP was implemented rise to the level of a well-resourced and operated Participatory Budgeting program, such as those studied in other Kenyan counties (Touchton and Wampler 2023). The perception that existing participatory planning interventions were not functional extended to both government participatory structures as well as to a sparse mixture of previous efforts by NGOs to generate participatory processes in a handful of wards. Notably, this assessment was also shared by government respondents, both in the executive and legislature. As a result, even if ward-level participatory planning had been previously introduced in form, a “genuine governance gap” remained in function.

When asked how the WDP process compared to existing participatory structures, there was a consistent perception among respondents that these pre-existing participatory planning structures were flawed and minimally effective. The WDP process was described in contrast as being qualitatively different, by focusing more intensively on building the capacity of WPCs as planning and advocacy institutions.Footnote 18 The WDP process was described as providing more resources and time for communities to diagnose local problems, deliberate on potential solutions, and employ an ongoing organizational platform for advocacy and accountability vis-a-vis government and NGOs.Footnote 19

This comparative perception was corroborated by most county-level government officials, ward administrators, and community members. Government officials and ward administrators were blunt about the limitations of Kenya’s constitutionally mandated public consultations on county budgets, which were often poorly resourced and short (half-day) meetings held at the sub-location level. Because these were open fora, discussions were often dominated by a “few loud voices.” There was also an unaddressed gap between the public’s understanding and the relatively technical budgeting discussions taking place, with little or no training on how to diagnose their community problems. Following the PB process, communities typically felt that public money was being misused or stolen due to either corruption or ineptitude. However, they were unaware of any clear accountability mechanism between initial participation and (non)received public goods.Footnote 20

In comparison, the WDP focus on capacity building and ongoing support for both planning, advocacy, and accountability was described as “extremely important,”Footnote 21 “eye-opening,”Footnote 22 and “never before seen.”Footnote 23

“This plan [The WDP] is more detailed and has a wider base of voices [compared to the previous participatory process]. Thus, more problems are being heard from [geographic] areas which, before, were not heard.”

37 - Isiolo, Government, Male, Adult

An exception illustrates that overlapping participatory function is a significant concern, whereas this issue is not a problem for participation in form alone, as indicated earlier. In a particular Garissa County ward, a project by the World Bank (WB) had implemented a participatory planning initiative for infrastructure development. This case was one of the few instances where the Ward Development Plan (WDP) process was redundant across multiple functions. Notably, both programs were well-funded and well-attended, establishing similar committee structures with several overlapping members, to the extent that the chairperson and secretary for the two institutions were the same individuals, merely swapping roles. Both initiatives mapped local requirements through participatory assessment and devised a plan for prioritizing public funding, leading to the identification of predictably similar priorities.

Interestingly, although the programs were redundant and the overlap was clearly inefficient, they were not regarded as interchangeable. Instead, they were viewed as having distinct strengths and weaknesses. When asked to compare the two, one respondent favored the World Bank model due to its assurance of direct funding.Footnote 24 Conversely, another respondent believed that the WDP program was more robust, as it cultivated a general capacity to advocate to both the government and non-governmental organizations (NGOs), while the benefits from the World Bank program were contingent on time-bound funding.

“As a community, we have woken up from slumber… In fact, [before the WDP] we were not aware that we were able to do this [advocate to the government]. We realized that we have to engage partners to lobby for development. This is a very good aspect. The World Bank [in comparison] doesn't have much training. It is very few [sic].”

4 - Garissa, WPC, Male, Adult

Meso: Triggering Causal Mechanisms: Interactions Between Design and Context

This section describes how specific causal mechanisms were or were not triggered due to an interplay between the rule-based design of the WDP and the contextual factors where the program was implemented (Fig. 3). This discussion demonstrates the importance of participatory planning intervention designers looking “under the hood” of programs to analytically identify the conditional mediators needed for designs to function in a given context. These relationships create “if… then…” logic that explain how the rules that define a given participatory planning intervention do or do not provide expected outcomes such as accountability or empowerment.

Fig. 3
figure 3

Meso-level: rule-based institutional design aligned with context

Pre-existing Civil Society and the Conditional Success of Delegated Representation

The delegated representation model used in the WDP relied heavily on pre-existing dense civil society at the local level to operate as intended. The selection process was typically a two-stage selection of candidates followed by a final selection process of either open (non-secret ballot) votingFootnote 25 or consensus-seeking consultation.Footnote 26 The multi-stage process allowed communities to select delegates with strong civic backgrounds and a record of community service. Community meetings were also well attended due to an established tradition of holding large-scale and open meetings at the village level, called barazas.

Due to the public selection process, WPC members were generally well known within their constituent communities, allowing them to convincingly represent the community to government and NGO actors. The vast majority of WPC members had some previous experience in a civic service role (e.g., community health volunteer, elder groups, economic association).Footnote 27 The Ward Administrator and other diverse ex officio government officials felt that the WPC members were key channels to the “grassroots,” implying that the selection process had produced individuals or adequate stature to both represent and relay information to the community-level.Footnote 28

One caveat to this finding concerns quotas for women, youth, and individuals with disabilities on the WPC. As these quotas were drawn from populations historically holding less power within the context, we find suggestive evidence indicating that these members may encounter challenges in fulfilling their WPC responsibilities in certain cases. For example, in more conservative regions, a female WPC member might not have an equal say during discussions,Footnote 29 while individuals with lower social influence in their home communities might struggle to gather the same attendance for community gatherings (barazas) compared to someone with a higher social standing.Footnote 30 Particularly, young women seemed to have the lowest pre-existing social status and authority. These young women seldom undertook advocacy roles and appeared to have moderately less awareness of all WPC activities, leading to less detailed and more second-hand accounts of WPC functions in comparison to their male counterparts and adult women on the same WPC. Footnote 31 However, female youth emphasized the significance of their role within their community, occasionally in terms of strengthening more equitable norms and most often for the valuable experience they gained. This suggests that quotas should be accompanied by additional support for identity groups with lower power when placed in leadership positions.

Societal Factors and the Conditional Operation of Informal Accountability

Another contextual factor which assisted the WDP design to function was a high pre-existing level of social cohesion at the village/village cluster and the ward levels. While intra-ethnic (i.e., clan-based) political competition is certainly possible within Kenya, ethnic identity has long been a more salient political cleavage in Kenya. In accordance with this, homogeneity of primary ethnic identity and, relatedly, high social interconnectedness due to geographic proximity was consistently noted as a reason for low conflict over resources within the WPC.Footnote 32

Ethnic homogeneity and social cohesion limited conflict within the ward level due to the threat of social sanction. Because most wards were homogenous (10 out of 12 included in our sample) with high social interconnection, committee members reported feeling that projects should go to those with the most need or else to areas where investment would have the highest marginal impact. When directly asked why people did not selfishly advocate for their own village (cluster) above others, respondents cited that it would be “shameful” to be too blatantly parochial during the WDP process, implying that social sanctions would be relevant between villages. Because of this, deliberations on priorities were rarely confrontational.Footnote 33

“Because we are the same clan and are all attending [the WPC meeting]… it would be seen as shameful. People would say ‘you are ego-centric; you love yourself too much’.”

51 - Isiolo, WPC, Male, Elder

There was no mention of mobilization across social divides or resulting conflict within the ward.Footnote 34 Similarly, we found no instances where the WDP process was not endorsed by all WPC members interviewed nor did government officials or implementing staff identify any instances. However, there were multiple accounts of inter-ethnic conflict, often occurring between county or national borders.Footnote 35

The second role of social cohesion, when combined with local embeddedness, was in providing strong accountability between the WPC member and their constituent village. As noted, the village/village clusters which WPC members were drawn from were typically homogeneous in both ethnicity and clan/sub-tribe. As a result, there was a lower threat of WPC members acting parochially within their village cluster. This accountability operated through the WPC member interfacing regularly with the wider population. This extended beyond the official public meetings required in the WDP intervention. Some WPC members chose to call supplemental meetings with community members at the village (cluster) to gather opinions, share what they learned in training, and/or update the community about the progress of the WDP.Footnote 36 Others reported naturally encountering and updating their fellow community members as they encountered them in day-to-day life or pre-existing civic roles, at which time community members requested updates about the program and shared relevant problems.Footnote 37

Embeddedness as a source of accountability has a very different logic than relying on elections or other forms of formal oversight (Tsai 2007). Though the two do not necessarily compete, our interviews suggest that embeddedness was more important than the threat of losing the WPC position in the next selection round. Instead of selection being an opportunity to punish WPC members, we find stronger evidence that it was an opportunity to select WPC members with intrinsic qualities of embeddedness and a trustworthy reputation. No WPC member listed campaigning on a specific outcome or policy position as an important factor in their election vis-a-vis another candidate. WPC members predominantly emphasized that they campaigned and were elected on personal characteristics such as being trustworthy or previous examples of civic action.Footnote 38 The capacity to advocate was another important factor, and it seemed that communities tried to maximize on the combination of both, though trustworthiness was more commonly cited.Footnote 39

Threat of losing their WPC position was never mentioned as an accountability mechanism. As an unpaid position with a relatively high time commitment, none voiced concern about losing their position, and most were not sure when the next selection would be. In one of the few cases where a new WPC election had already taken place, an outgoing member was nonchalant about the loss and reported being happy that a younger, more energetic person was replacing him.Footnote 40

The other institutionalized accountability mechanism—public validation of the WDP document among constituent communities—also seemed to play a subordinate role to social embeddedness. While reportedly a useful opportunity for formally communicating the draft WDP to the community, we find no instances where the public was displeased with the proposed plan during validation or resulted in a serious re-work of the draft document. Of course, the mandated public review may operate as a passive accountability mechanism, but no WPC member framed it as such, and it seemed to complement, rather than drive, ongoing communication and accountability, based on embeddedness and social cohesion.

A caveat to this section is that the WDP process had extremely limited funds flowing directly through it. If this changed, we would expect the current accountability measures—which rely heavily on embeddedness and pre-existing social cohesion—to be strained. In this case, more institutionalized accountability processes may become important. The power of the WDP was in its ability to legitimately articulate the desire of the ward, but this power was purely consultative, and WPC members did not control any meaningful development budget. While the WDP in its current form was highly trusted, if the WDP was formally tied to a governmental or NGO budget, the risk of conflict or capture would increase substantially.Footnote 41

Capture may come from multiple sources. First, political interests—local chiefs or county officials—may try to influence the selection processes. Again, we find only a handful of instances where this was already happening: one instance where a chief’s wife attempted to run for the WPCFootnote 42 and another where local chiefs attempted to circumvent the selection process entirely and select their own candidates.Footnote 43 While it is likely that more attempts exist than uncovered, in all reported instances, an implementer was able to stop these efforts through negotiation and leveraging their status as a staff of an INGO with strong monitoring mechanisms. Second, the low personal benefit of the position limited any adverse selection for unscrupulous actors to vie for WPC membership. This selection effect (Mansbridge 2009) seemed to maximize the pro-social qualities of candidates and limit capture internally to the WPC, building on the social cohesion and embeddedness discussed above.

Binding Decisions and the Conditional Risk of Government Capture

The design of the WDP process strategically emphasized its relationship to government and NGOs to offset the non-binding nature of that relationship. Unlike Participatory Budgeting, which is explicitly meant to inform how governments allocate public spending, or CDD which devolves a budget directly to the participatory committee, the WDP process was not binding on any significant government or NGO budget. While WPCs were informally aligned with the national climate law, which mandated ward-level committees, no government had formally recognized the WPCs, and all stakeholders acknowledged that the WDP process was associated with the government through good faith without formal control over county spending or the spending of other NGOs in the ward.Footnote 44

Because of this, the WPC relied entirely on persuasion and advocacy rather than institutional rules to shift funding. In response, the WDP process placed a strong emphasis on engaging the county government throughout its various stages. This integration was strategic, allowing them to increase buy-in from county officials by establishing or strengthening ties between WPC members and the county. This integration ensured that the WDP process was perceived as legitimate from the county’s perspective.Footnote 45 This strategy was crucial for the success of the model, since the WDP model could only influence, not coerce, decision-makers to voluntarily follow the priorities outlined in the WDP document.

Closer partnership with the county has the potential to guide development spending in two ways. First, it influenced the county’s allocation of the development budget, and second, county government representatives who were aware of the WDP document championed its use by other NGOs investing in the ward. In this, the county officials suggested the WDP as a coordinating document, alongside official government plans. County officials across various departments were interested in using the WDP document in this way. Yet their ability to force other NGOs to follow the WDP was limited, because the WDP document was not yet formally acknowledged by the county. As such, they could only suggest other NGOs take it up. That said, multiple county officials reportedly encouraged NGOs to align with the document rather than conduct an entirely separate participatory planning process.Footnote 46 In two cases, external NGO validated the WDP document through a series of community meetings rather than conducting an entirely new participatory process, suggesting the legitimacy and quality of WDP.Footnote 47

This integration of county officials was emphasized throughout all phases of the project.Footnote 48 In the planning stage, ground-level implementers worked through the county offices and Ward Administrators to select wards and access village-level communities. During implementation and training, county and national staff were invited to provide technical training from various relevant departments. In addition, a wide range of government officials, including the WA, MCA, local chiefs, and other county officials, were invited to act as ex officio members, attending meetings and training throughout the entire process. Though attendance varied widely between wards and individuals, at least a handful of ex officio members across all wards reportedly attended meetings and training.

The inclusion of government officials was a conscious choice in the WDP model and contrasted sharply with the above-mentioned participatory program funded by the World Bank (WB) in Garissa County and implemented in an overlapping ward. There, the WB, according to a government official, partnered closely with the county until implementation began but refused to allow the county to either monitor or be involved with the process once implementation started.Footnote 49 Because of this, the county had little knowledge of what the project was doing, whether it aligned with the county’s development plans, or even whether the committee generated was legitimately formed. The stated reason for this, again according to a government official directly familiar with the case, was that the WB did not want the government to interfere, with the strong implication that the WB was concerned about corruption or some other political influence. The relationship between the two actors deteriorated, and the government official did not trust them. However, the county could not refuse to allow the WB to operate, as it had such significant funding.

This seems to be a genuine trade-off with regard to government involvement. Greater government involvement in participatory processes increases the ability for participatory plans to be coordinated with government plans and for participatory institutions to engage with and advocate to the government, yet also increases the risk of capture or political interference. Within the WDP process, it seems that this risk was relatively limited because there was no external money flowing through the WPC and the county had full discretion to follow or depart from the WDP priorities in its own spending. As such, the inclusion of the government had limited downside, while the upside (greater influence and coordination) was core to the program’s success. Besides the direct impact on program outcomes, the inclusion of government in the WDP process reinforced the connection between elected, formal government representatives and their constituents.

While including government seems beneficial overall, as discussed above, the direct inclusion of ex officio members does represent a risk. As discussed above, there were a handful of examples where ex officio political actors tried to influence selection or funding. Another concern, however, is that ex officio members may manipulate the deliberative process if invited to meetings. One elected government official interviewed, who served an ex officio role, felt that the less politically savvy WPC members may not even notice they were being manipulated or influenced, even suggesting that he himself had done this.Footnote 50 If true, this would be difficult for our approach to detect, via interviewing WPC members. We were not able to trace any specific instances where this influence may have impacted the final WDP document, much less shifted funding, and none of the respondents felt this influence was overwhelming or warranted removing the ex officio members.

Two counter-examples, where high government integration failed to produce results, showed both the importance and limitations of including government officials. The first example suggests how important this integration is, as demonstrated when the program fails to involve a key “link” between the government and the WDP process: the Ward Administrator. This happened in a single ward in Isiolo county, where a ward official either did not hear aboutFootnote 51 (or was non-responsive to,Footnote 52 accounts differed) the WDP process in his ward. In this case, the ward offical did not attend any meetings, was not aware of the program’s existence in any meaningful way, and therefore played no role as an intermediary between the WPC and the county governmentin contrast to many other wards where the WA played this role. Because of this, WPC members reported struggling to either establish communication with or influence the county government generally.Footnote 53 They provided the one poignant example that the WPC and community were not consulted when the county government put in a new borehole, leading to it being established in a community-designated strategic drought reserve and thus ruining the resilience efforts of the community (as it attracted pastoralist foot traffic to the water point).Footnote 54

The second example shows the limitation of integration and influence alone. For influence to be effective, the local political economy must be reasonably responsive to advocacy or requests. In cases where the county government was largely non-aligned with wards, being more integrated with the government was insufficient to materially change policies. While multiple WPCs felt their advocacy was being rebuffed, the most dramatic example was of a ward populated by a small ethnic minority within the county. In this case, the WPC and Ward Administrator (of the same ethnic minority) felt that the integration with the county increased their ability to advocate, by providing training about how to make political demands, but their advocacy was ineffective, which they attributed to the deeply ingrained preference of the county government for people from the majority ethnic group in the county.Footnote 55

Empowerment and Descriptive Representation

This section examines the interplay of context and program design for empowerment of marginalized groups within participatory planning. We briefly consider empowerment via three distinct pathways: (1) via descriptive representation of marginalized groups on the WPC via quotas and inclusion in public meetings, specifically women, youth, and individuals with disabilities; (2) via geographic distribution of the WPC members across the wider ward, in which no village/village cluster could monopolize resources; and (3) broad-based empowerment vis-a-vis county government.

Regarding the descriptive representation of marginalized groups, we find the pathway was only weakly triggered, as the quotas were too modest to substantially deviate from discriminatory social norms related to marginalized group participation. In contrast, the mandated geographic distribution of WPC members across wards did seem to shift the logic of parochial interest, where political actors of all stripes were expected to act parochially on behalf of their home communities.

Perhaps the largest shift in empowerment was of the community vis-a-vis the county government and NGOs. Rather than being passive recipients of development aid, a common theme among respondents was that they felt empowered to engage due to new capacities and understanding of both development and governmental processes. This empowerment is akin to the empowerment via enhanced capabilities and collective as described by Amartya Sen in Development as Freedom, in which citizens are able to assert control over the development processes surrounding them (Evans 2002).

The WDP process had fairly low quotas for identity groups, including women (two of 11), youth (two of 11), and disabled (one of 11). We find that these low quotas were likely to have only a marginal impact on empowering these groups, in part because they did not deviate substantially from existing social norms. Across all wards and counties, respondents —including women and youth—felt that their roles had shifted substantially towards more inclusion over the past two decades.Footnote 56 At present, it was common for NGOs to require their participation without much public comment or concern. Rather, implementing staff suggested that many families vied for female and youth family members to join the program, in hope of some compensation or benefit.Footnote 57 Hypothetically, we would expect that a more extreme quota, say 50% of all WPC members must be female youth, may represent a more substantial deviation from social norms, challenging male-dominated societies.

The operation of quotas for participation was marginal and seemingly contingent on the degree of deviation from local social norms. In more conservative areas, widening participation was a larger normative shift.Footnote 58 We do find that women and youth used their role to advocate for their representative groups, in some cases, resulting in specific public goods being included in a development plan, such as funding hairdressing classes for women.Footnote 59 Yet, these group-specific projects were consistently described by marginalized groups as secondary to more severe population needs, such as water, economic support, and education.Footnote 60

A central element in the WDP process was the geographic dispersion of WPC representatives across the ward. This helped ensure that WPC members were embedded within each village area. Yet, this does not fully solve a common problem in Kenya, which is that each politician engages in clientelistic behavior on behalf of their constituencies. Social cohesion at the ward level partially solves this, but we also find that equal geographic distribution of WPC members across the ward ensured that no single village was able to dominate.

Since the 11–15 members are drawn from across the ward, each WPC member represents a distinct locale, collectively covering the entire ward. WPC members, even from minority groups, felt that this geographic dispersion was sufficient for limiting the domination of a single village or cluster of villages.Footnote 61 Respondents suggested that this distributed design hinders attempts to co-opt the body, since they represent a broader array of geographic interests across the entire ward.

Q: How will the WPC avoid bias in the distribution of resources?

A: “Every location is represented, so there is no way that one location will be able to capture the money.”

8 - Garissa, WPC, Male, Adult

In one illustrative instance, the WPC design’s emphasis on representation for each village across the ward was able to mitigate an ongoing pattern of parochial, clientelist treatment. An MCA had previously directed a disproportionate amount of funding to his home village within the ward. After the WDP process, this behavior became less tenable, as it could be cross-compared against publicly stated priorities of the entire ward, including the MCA’s hometown. As a result, money was shared more fairly among the ward afterward, and within the WDP document, the MCA’s hometown was moderately less included, simply because it had more previous investments and therefore less objective need.Footnote 62

An important form of empowerment highlighted by many respondents was the recognition of the community as a co-equal agent within the development process. While physical goods were important, respondents emphasized the qualitative feeling of empowerment they experienced, especially the ability to understand and advocate within decision-making processes for public goods. Empowerment was not only valued for its ability to attract resources. Interviewees repeatedly referenced empowerment as valuable on the grounds of human dignity and increased self-efficacy.Footnote 63

This contrasted with previous experiences of empowerment. Communities expressed recurring frustration with county government and NGOs failing to consult them seriously or to build their capacity to understand the development processes. Rather, their experience was as passive recipients. Notably, this shift was also remarked on by the governments to whom they were presenting demands.Footnote 64

Empowerment vis-a-vis county government and NGOs stemmed from a combination of high-quality training and capacity strengthening—which emphasized their rights as citizens to demand public services and how to access them, by, for example, writing a proposal. Empowerment also came via the production of a high-quality WDP document, which was professionally formatted and structured to be legible within the professional government and NGO development space. Copies often showed signs of use and some interviewees demonstrated keen familiarity with its content. One interviewee also referenced supplementary ‘tracking” documents, showing which projects had or had not been funded. The program provided both the formatting and publishing of these documents, and WPC members felt it signaled a degree of formality and legitimacy to their demands.Footnote 65

A caveat to both of these senses of empowerment is that they may be vulnerable to disillusionment if governments and NGOs are unresponsive. Our research did not find any instances of this, yet we only spoke with WPC members willing to sit for an interview and who therefore were still minimally engaged with the program and, likely, hopeful of a positive outcome. A handful of WPC members did express exasperation that so little money had yet flowed to their ward after 3 + years. This could potentially morph into broader disillusionment. While we cannot provide any detailed answer regarding the trajectory of perceived empowerment in the face of perceived failure, it is an important point of further investigation, especially in situations where participatory planning is non-binding, such as our case.

Micro: Discretionary Adaptation and Managing for Motivation

The final section of our findings moves away from specific program elements and turns to the process of implementation (Fig. 4). We find that the WDP implementation process was reliant on the ability of high-capacity local implementers to operate semi-autonomously, with very limited direct oversight in the field. This autonomy allowed implementers to use strategic discretion regarding how to best implement the program’s broad design features. We briefly identify three aspects of this discretion: (1) shifting design features to be locally recognizable and legitimate, (2) protecting the program’s causal logic from political interference, and (3) exerting effort to ensure the program’s causal logic is effective.

Fig. 4
figure 4

Micro-level: discretionary adaptation of ground-level implementers

A key discretionary feature of project implementation is making a program’s abstract components locally recognizable and legitimate to reinforce compliance and buy-in while avoiding backlash or “friction” with the community (Pain 2018). Doing this well requires a detailed understanding of the idiosyncratic local political and social context—or “metis” using Scott’s (1998) terminology of practical knowledge rooted in a specific context—which ground-level implementers are best positioned to bring to bear.

A recurring example of this discretionary adaptation being applied concerns the selection of WPC members. For the program’s theory of change, it is crucial that this process be perceived as open and fair within the community. Yet, at the community-level context in Kenya, what a legitimate and fair selection process looks like differs by region and ethnicity. Ground-level implementers were sensitive to these differences and strategically deployed the process most familiar to the community. Thus, within ethnic Turkana communities (across both Turkana County and Isiolo County), we found that, after a deliberative and open process for selecting candidates, queue voting was used almost exclusively to select the successful candidates.Footnote 66 However, within Buran and Somali communities, deliberation would select the candidates, followed by choosing the final member by consensus-seeking consultation.Footnote 67

The institutional choice set went beyond the “toolkit” choice set provided by Mercy Corps for WPC selection. The toolkit did provide options for whether to have single or two-stage selections and encouraged candidates to leave the room, but it did not address the specific decision-making mechanism by which individuals were selected. In this case, it seems that implementers were not only using discretion to select between pre-specified options but were actively deciding between options not addressed in the manual.

Another example of a trade-off between local legitimacy and international “best practice” was a wide preference for open voting rather than secret ballot. We saw only one use of a secret ballot across all 12 wards. We found that a secret ballot election process was, surprisingly, considered less legitimate and trustworthy than public and observable selection mechanisms. This can explain why implementers generally did not insist on it, as they were concerned with legitimating the WPC in the eyes of the community while avoiding friction. The explanation given by WPC members is that secret ballot elections are explicitly non-transparent since choices are hidden and, by extension, the process is less trustworthy.

Implementers use similar local knowledge to ensure legitimacy and recognizability when calling meetings (e.g., barazas), strategically engaging and getting buy-in from important elder, religious leaders, and local business groups, and also in navigating how to ensure the inclusion of marginalized groups such as youth, women, or the disabled.

Protecting the Program’s Causal Logic from Political Interference

A second basic category of discretion is in strategically protecting the program from political factors which may undermine the theorized causal operation of the program. This requires judgment and discretion on the part of the implementing team.

For example, in one ward, the implementer changed the selection mechanisms to secret ballot when a local chief’s wife stepped forward as a candidate. In doing this, the implementer again went beyond the formal program design, which did not include secret ballot. This is also the only instance we find secret ballots being used, further reinforcing that it was the individual initiative of the ground-level implementer. Also as discussed above, ground-level implementers had to block direct attempts by local power holders to interfere with the process by installing their own candidates. In this, they could not follow any set protocol with managing the social relation. The implementer also bore the social risk of displeasing powerful political actors within the community.Footnote 68 In a more prophylactic example, when initially calling local leaders to a meeting to plan how to gather the community in a baraza for sensitization about the project and selection processes, ground-level implementers were sure to also invite civil society organizations, since this would limit the ability of local leaders to act as sole gatekeepers for who attended the event, and the possibility that they would “stack” these events with political allies.Footnote 69

Intrinsic Motivation to Accomplish the Program’s Causal Logic

The broadest, yet crucial, category for discretionary action concerns translating an abstract design of participatory planning into real-world action. Participatory planning includes multi-round interactions with a dizzying array of community, government, and civil society stakeholders, each representing some degree of opportunity or threat for the program. However, interaction is costly in terms of time and effort for the implementer, leading to enormous latitude in terms of what “implementation” means (i.e., one hour spent recruiting can vary on the intensive margin and the creative tactics used to recruit).

Accountability and oversight, no matter how granular, are unlikely to replace the value of intrinsic motivation, combined with authorization from management to make discretionary decisions. Take the basic element of community meetings within participatory planning. These meetings lie at the heart of participatory processes—and it is the deliberative, public, and transparent nature of these events which provides legitimacy, critical information, and an opportunity for broad empowerment. Yet, many of the most causally important elements of these meetings cannot be mandated, measured, or observed.

An event as distinct as attendance can be measured and authenticated; yet, this will omit pivotal qualitative aspects of these meetings: are descriptively marginalized groups representative or selected via gatekeepers? Once arriving in the space, are there political or normative dynamics which would prevent marginalized groups from sharing their “true” opinions, such as fear of retribution or restrictive norms? These qualitative questions are irreducible to monitoring designs and are not limited to empowerment. For instance, during training, selecting the appropriate degree of sophistication requires localized judgment in both planning and in-the-moment adaptation.

Though the management of the WDP process was not an initial focus of our research, a few characteristics of internal management seemed key for fostering intrinsic, mission-driven motivation (Honig 2022). While there were implementation guidelines and toolkits guiding the development of WDPs, implementing staff at the local level—those who actually went out to the villages on a regular basis—reported being given significant discretion and agency to make in-the-field adaptations and even deviations from the formal design as they confronted local realities.

This is evidenced by the examples above, but the more difficult to capture are the myriad micro-decisions involved in implementing community meetings and the broader participatory planning process—e.g., dealing with recalcitrant chiefs or knowing how to best approach civil society groups. Rather than managing through top-down compliance alone, there was a flow of communication between staff and management. These communications were largely consultative and problem-oriented and did not seem to impose a rigid formula for micro-level decisions. Aligning with current research among managing for motivation rather than by top-down control, this style protected the intrinsic motivation of field-level implementers, and crucially, this motivation was combined with authorization from the management structure to make field-level decisions strategically allowed implementers to navigate by judgment (Honig 2018; 2022).

Conclusion and Implications

We set out to better explain when and how participatory planning models function to empower communities by providing inclusive spaces for broad-based participation and deliberation, as a distinct question from whether empowerment, once achieved, leads to downstream development outcomes (Woolcock 2019; Gibson and Woolcock 2008). Our top-line findings stand in partial contrast to the recent evaluation literature, which generally adopts a pessimistic stance, suggesting that participatory planning, as a class of interventions, does not “work” for producing social change or, at best, yields mixed results (Casey 2018; White et al. 2018).

Through mapping sub-national applications of one intervention, we find that participatory planning’s success is possible, even in adverse contexts. However, realizing this possibility is mediated by three levels of conditionality underappreciated by both practitioners and academics. These are as follows: (1) variation in local formal and informal governance systems and the challenge of avoiding redundancy; (2) a careful alignment of rule-based design with contextual mediators (e.g., structure and density of social relations); and (3) the use of discretion and adaptation by ground-level implementers as they translate an abstract intervention design into real-world action.

These findings introduce additional complexity for scaling participatory planning processes. At the macro-level, the extent and nature of governance gaps can be expected to vary significantly across national and local contexts, and, especially in respect to informal institutions and other participatory institutions, adding new institutions may have diminished or negative effects. At the meso-level, scaling and transferring the model to new contexts should require a significant amount of diagnostic and analytical effort to understand how participatory rules map onto contextual factors such as identity, social cohesion, and economic inequality.

Finally, at the micro-level, the success of implementation is dependent on the capacity of field implementers to use discretion, acting as “custodians” of a transactionally intensive implementation process meant to empower communities through inclusive participation. Even if the rule-based design is excellently matched to the national context, implementation still should intentionally allow for adaptation at the community level, where ground-level implementers make trade-offs and adjustments according to local enabling or blocking factors—for example, deciding which local norms and institutions to accommodate or resist (Gibson and Woolcock 2008).

This account of intentional input variability across all levels is in contrast with the current literature’s emphasis on attempting to capture whether participation “works” in general. These findings also cut against a tendency in both research and programming to increase standardization and compliance for accountability to donors. If we are correct, this conditional account will allow researchers and practitioners to escape the inevitable conclusion that aggregated evidence is weak or mixed, due to unacknowledged heterogeneity in input type and dosage between and within country-level applications. This adoption would take the form of rethinking “best practices” and black-box experimental evaluations, which stress input homogeneity, and towards a greater interest in understanding how to best customize program design and implementation, contingent on the realities of each context.

Taken together, our findings suggest the need for a re-orientation of research on participatory planning. Evaluating participatory programs using experimental and quasi-experimental designs may be valuable for retrospectively determining the value of money spent, but—unless accompanied by a mechanistic and qualitative account explaining what occurs in the “black box”—it is of limited prospective value for gauging when and whether a program will produce similar outcomes in another context (Andrews et al. 2017; Cartwright et al. 2020; Cartwright and Hardie 2012; Woolcock 2013).

Additionally, evaluation efforts which require standardization of inputs for causal identification may themselves cause a program to limit the discretion of implementing agents or increase top-down accountability structures, thereby adversely affecting the quality of implementation.

We believe there is immense opportunity for future research and reflective practice to further develop, explore, and refine the theoretical approach we have started to build in this study. As a simple illustration, the design of the WDP intervention may need to be adjusted in the course of “scaling” the project within Kenya—a likely proposition given that numerous county legislatures are considering expansion bills. One major change proposed in the legislation is a specified budget in the form of a “Ward Development Fund,” which would be allocated to both support the operation of WPCs and fulfill the public projects identified across wards.

The introduction of a specified government funding source would substantially shift the WDP closer to Participatory Budgeting, as outlined in Table 2 above. The government would take a larger role as a facilitator, and this would introduce the government budget as a specified source of income. In this instance, our findings would suggest a re-evaluation of whether the embeddedness of elected WPC members would be sufficient to avoid capture, given the greater moral hazard of directly impacting government budgets. Additionally, the aggressive inclusion of government actors (e.g., through designated ex officio members encouraged to attend all meetings) operated as a pathway of influence for the WDP program. However, if the government is operating the program which influences a county budget, program designers may instead ask how to ensure adequate space for communities to operate without interference or influence.

These questions into effectively managing the WDP’s planned shift from an NGO-led pilot to an official, government-operated institution also extend to broader discussions in policy and programming concerning CDD and Participatory Budgeting PB. The answers to these questions are not universally applicable; they rely instead on specific national and sub-national circumstances. Both the risks of capture and the potential for enhanced accountability, stemming from the involvement of government actors, can be evaluated on a case-by-case basis. The contingency examined in this study pertains solely to the meso-level of institutional design, but comparable questions and analyses must be considered at the micro-level (e.g., safeguarding the process of recruiting ground-level implementers and intensifying efforts to ensure mission-oriented drive) and macro-level (e.g., ensuring that legislative bills do not amend the institutional design to the extent that the governance gap is no longer addressed).

As an alternative path of inquiry, research may consider the functioning of specific mechanisms within a broader participatory planning program, such as the comparison between voting and consensus decision-making. One approach to accomplish this might involve randomizing specific components while limiting local discretion. Beath et al.’s (2016) randomization of the selection process for participatory committee members serves as an example. While we question the theoretical value of the larger experiment, which measures the impact of the entire program, this alternative focus on comprehending a specific design element appears to be more feasible theoretically, particularly when supplemented by qualitative research. Such qualitative research could provide insights into how various selection processes lead to differences in downstream outputs and outcomes.

Another potentially fertile area for research is on how a program theory of change is communicated to field-level implementers and how this understanding guides strategic decision-making. Currently, program theories of change are positioned at an exceedingly abstract level, predominantly aimed at donors and management. This approach seems logical, especially when program design is largely determined during the proposal stage (where the funder decides whether to proceed) and by the management structure of NGOs (which determines how to adapt a model to, usually, a country-specific context). However, if we recognize that field-level implementers must make strategic decisions aimed at maximizing the activation of causal processes assumed by the theory of change, their comprehensive and close grasp of the theory becomes significantly more crucial. In essence, they are effectively making design choices during day-to-day implementation.

The utilization of intuitive theory by these implementers and the process of integrating theory into strategic decision-making are largely unrecognized (owing to the assumption of mechanistic, standardized implementation). Consequently, it is poorly comprehended and represents a promising area for further investigation.